The Art of Analysis

It is common knowledge that in order to improve your game you need to analyze chess games, most importantly your own chess games. But how many people know how to do it? Today, I am going to guide you through the game and ask you to do the chess analysis of certain positions. Try to do it in the following steps.

1. Analyze the position mentally, just like you will do it over the board, but take your time this time!

2. After that open the analysis board and check your variations on the board, see if you can find better moves.

3. Finally, go with the chess engine to check your analysis and see what you missed.

When you are done with all preparation steps go on and take a follow-up quiz, if you did your analysis properly you should be able to solve it!

How much analysis you do in the preparation for the quiz is totally up to you. You can make it quick and casual (which is the most likely scenario if you want to do an analysis of your own game) or you can try to go deeper and uncover the hidden secrets of the position. Most of the beauty of chess is hidden in variations. By doing a deeper analysis you will uncover the truth of the position and get a new appreciation of the depth of the game.

But no matter how deep you will do your analysis and at what level you are right now by doing it you will improve your game!
 

Some tips for the technique of chess analysis.

1. Before considering any variations look for ideas (tactical and positional) . What are the weakest points in the position, can they be exploited?

2. Open the analysis board and begin to enter variations (but do not activate the chess engine yet) ! When you find a refutation of the move track your moves back and find a better move. Pay special attention to very first moves in a variation; they are the most important.

3. When you are satisfied with your variation tree begin to check it with the chess engine. See where you went wrong and figure out what was the mistake.

4. For a true understanding of a position you need to understand transitions to middlegame and to the endgame. Do you understand if a particular transition to a particular endgame is beneficial or not?

5. Always look for variations that might turn the game to your advantage. If you do not see an easy refutation your opponent will also have difficulty finding it. Do not stop your analysis until you find such variations.

6. It takes a lot of time and effort to make a real breakthrough, but it is very satisfying when you can do it.

7. Use mostly stockfish for tactical positions and lc0 (Alpha Zero clone, but stronger) for quieter positions. Always try to find a reason for a particular engine move.

8. Even when you are analyzing with the chess engine you are analyzing against the chess engine. You want to improve on the chess engine's 'best line'. The greatest danger of such an analysis is to begin to follow engine recommendations blindly.

Some 'Pro' tips for chess engine analysis.

1. You can run stockfish and lc0 at the same time (especially if you have a good GPU) . Set up lc0 to use 2 cores and give the rest to the stockfish. Always set up stockfish to show one PV (principal variation) . If you set it to say 3 you will slow it down three times, but set up lc0 to 7-10 variations (it will not slow it down due to different architecture) .

2. Most chess UI (User Interface) will have options to include or exclude variations the engine will analyze. You should never set up an engine to analyze more than one line unless it is an MCTS engine (like lc0 or Komodo MCTS) ! Instead, start the engine with one PV and later restart analysis with the first line excluded, then two best lines excluded, and so on. This way you will get multi PV, but the engine will increase its depth faster in each variation and you are not limited to how many different lines you can analyze in this way.

3. Pay special attention to the moves at the root of your tree (always use the above technique at the root) ! If you missed an important move at the root of the tree the rest of the analysis becomes obsolete.

4. If you let lc0 run for a long time it will eat all your RAM memory, you can set up the limit on how much memory it can use. Give the rest to stockfish. The more RAM stockfish will have the more useful it will be for 'backpropagation' analysis (see next tip about backpropagation) .

5. A very important technique with regular alpha-beta engines (like stockfish) is backpropagation analysis (but it does not work yet with MCTS engines) . When you move forward along the PV your move evaluation might change significantly. If this happens go to the previous move (backpropagate) . Most likely the engine will choose a different line with better evaluation. If it does not happen, backpropagate again until evaluation will improve. This is one of the most useful tricks in chess engine analysis.

Question: How you can apply what you learn to the analysis of your own games? In particular, how can you determine a critical moment in the game?

Answer: The main idea of training is to teach you how to do chess analysis by doing chess analysis. What I mean by that is, as to learn chess tactics you need to practice tactics exercises, to learn chess analysis you need to practice analyzing games.

When you want to apply it to the analysis of your games I would suggest the following steps:

1. Run engine analysis which will create an annotation to your game.

2. Begin to analyze your game just before a new annotation is created.

3. Go through steps I mentioned in the introduction chapter (again how deep you want to go depends on how much time you can spend on the analysis) .

4. By comparing your analysis with engine annotation you can evaluate the quality of your own analysis.

5. You can continue to analyze the 'critical' position (where the engine annotation was added) deeper with the chess engine, to see if you can improve variation or find alternatives.

By doing this often you will develop a better intuition for identifying which moments of the game are 'critical.'. This is the most difficult part of the OTB chess game: figuring out when you have a critical position on the board! It requires a lot of experience, but perhaps the following ideas might help:

1. Consider all 'irreversible' changes of a position as critical (pawn moves, exchanges of the pieces, sacrifices, etc.)

2. If you could not find an easy move for you or your opponent - this is a critical position.

3. Transitions are always critical, especially when there are different choices of what transition can take place (for example pawn endgame versus, rook and pawn endgame) .

4. Unfamiliar openings or variations in the opening.

Critical positions are critical (no pun intended) for the correct time management in the OTB game, but it is much easier to find them later in post-game analysis. By creating a habit of doing a post-game analysis after each game (even if it is blitz or rapid) you will learn to recognize when it is a critical position in the OTB game.

Question: For step # 7 ( Use mostly stockfish for tactical positions and lc0 ( Alpha Zero clone, but stronger ) for quieter positions. ), I'm wondering why I should do that! I would (perhaps, naively!) imagine that any level below super GM either engine would be more than sufficient.

Answer: Oh, this is a very interesting point. The engines evaluations can be very different and in some positions, you will trust more to one engine (tactical - stockfish) in some to another one (quiet and positional - lc0) . Also, lc0 variations are more understandable from a human perspective and more logical, more 'connected'. The stockfish evaluations are much more 'chaotic' and calculation dependent. In quiet positions, stockfish gives you almost random variations without any logic behind them.